The performance of ADSGA that is proposed for additional learning of RBFN. From left to right: density distributions of learned data when 2 parameters (K11,K33) are estimated using k NN (k = 2, 4, 8, and 16) and ADSGA, where the X-axis and the Y-axis represent logarithmic coordinates of data. 'Answer' represents coordinates of the optimal solution. Deepening colors represent the increasing density of existing data. The Grids represent 0.1-100 from the left-bottom. This results show that ADSGA applied RBFN searches both data-sparse areas and answer-adjacent areas simultaneously.